Alibaba has released an efficient open-weight model that it says rivals one of Anthropic's flagship systems, in another sign that Chinese and open models are narrowing the gap with the biggest proprietary AI platforms.
Qwen3.8-Flash-Next is a new open-weight AI model designed to deliver high-end performance more efficiently than traditional giant models. It has outperformed Anthropic's Claude Opus 4.6 Max on several benchmark performance tests, including for software engineering, coding, and scientific reasoning. The results are largely first-party benchmarks, however, so independent testing may better determine how well the model performs in real-world use.
Qwen's release comes as Chinese open-weight models gain wider adoption and increasingly challenge U.S. companies like OpenAI and Anthropic. CNBC reported this summer that the rise of those models has also drawn attention from U.S. policymakers concerned about the rapid overseas adoption of Chinese AI alternatives.
Qwen3.8-Flash-Next is multimodal, meaning it can handle more than just text, and uses a "mixture-of-experts" design. The model has 125 billion core parameters plus 51 billion additional n-gram embeddings, but activates only about six billion parameters for each token it processes.
The launch also comes as investors focus on whether Alibaba's heavy AI investment can translate into stronger cloud growth and profitability. In February 2025, the Chinese firm said it would commit 380 billion yuan (or about $53 billion) into AI and cloud infrastructure over the next three years.
Mizuho analysts recently said Alibaba's shift toward higher-margin AI demand and progress on in-house chips were making a path toward more than 20% long-term cloud margins increasingly visible. In an Aug. 20 client note, Mizuho maintained an "outperform" rating and $195 price target on BABA shares. Alibaba's stock currently trades around $120.